Skip to content
Author / Builder

Building the connective tissue for agents.

Research notes from operating across the provider ecosystem—published in the open, with the pricing and code to back them up.

Builder / Route Tools

Brian Sparker

I build infrastructure that lets agents choose tools without inheriting provider lock-in.

GitHub

I build route.tools, a router for the commodity infrastructure agents call: web search, scraping, document parsing, media generation, transcription, embeddings, and code sandboxes.

Building a router means living in every provider's docs, pricing pages, and failure modes at once. This journal is where I write that research down—honest comparisons, real workload math, and patterns that keep agents running when a vendor has a bad day.

The prices and quality scores come from the same open-source catalog the router reads. If something is wrong, open an issue. I would rather fix it than defend it.

Published field notes42 articles
01 Perplexity's search API in 2026: a guide for agent builders 02 How to audit what your AI agent actually spends per task 03 Kling vs Wan 2.2: the two hosted text-to-video APIs compared 04 Code sandbox security for AI agents: what actually matters 05 How to avoid vendor lock-in with embedding APIs 06 Transcription accuracy vs price: when cheap is fine and when it isn't 07 The xAI Grok image API: a practical guide for 2026 08 The 7 tools every AI agent needs (with real 2026 prices) 09 MCP vs function calling: when a protocol beats a schema 10 What an AI image actually costs in 2026: five API hosts compared 11 LlamaParse vs Reducto: which document parser fits your stack in 2026 12 Web search API latency: what actually matters for AI agents 13 BYOK economics: when bringing your own keys beats bundled pricing 14 ElevenLabs vs LMNT vs Deepgram Aura-2: the 3.3x TTS price spread 15 Spend caps for AI agents: budget safety before the loop runs away 16 Semantic vs keyword search: routing agent queries to the right tier 17 The 3 best embedding APIs in 2026: price, dimensions, and tradeoffs 18 Quality scores: how we rank API providers (and what the scores are not) 19 The 6 best web search APIs for AI agents in 2026 20 Scraping failures: blocks, empty bodies, and failover patterns that work 21 How tool routing cuts agent costs (and where it honestly doesn't) 22 Deepgram Nova-3 vs Groq Whisper: what the 6x price gap buys you 23 Prepaid credits vs subscriptions for AI APIs: why prepaid fits agents 24 The best video generation APIs in 2026: an honest short list 25 The AI agent infrastructure stack in 2026: who does what 26 Embedding API pricing breakdown: the per-token math that matters 27 Firecrawl for agent builders: scraping, search, and failover fit 28 Normalized APIs vs provider lock-in: what one schema really costs 29 FLUX image API hosts compared: same model, three different prices 30 The cheapest transcription API in 2026 (and when cheap is enough) 31 Exa neural search: a practical guide for agent builders 32 Circuit breakers for AI agents: stop retrying dead providers 33 The best text-to-speech APIs in 2026: Deepgram vs LMNT vs ElevenLabs 34 MCP servers explained for developers: local, remote, and why fewer is better 35 E2B vs Daytona: choosing a code sandbox API when the price is a tie 36 Understanding per-unit pricing for AI APIs: a billing literacy guide 37 The best document parsing APIs for RAG pipelines in 2026 38 Serper vs Brave Search API: reseller economics vs an independent index 39 Why AI agents need failover more than any software before them 40 Web scraping APIs compared: Jina vs Spider vs Firecrawl in 2026 41 OpenRouter for tools: where the analogy works and where it breaks 42 What is a tool router? The missing layer in the AI agent stack